Mild cognitive impairment (MCI) is common in older adults and represents a high-risk group for Alzheimer's disease (AD). Advances in biomarker assessment have improved the early diagnosis of clinical AD but medication trials in MCI have generally failed. New discoveries about brain plasticity in aging have led to the study of cognitive training as a treatment to improve cognitive abilities. Computerized Cognitive Training (CCT) provides an exciting opportunity and a new strategy to improve cognitive performance in MCI. CCT involves computerized cognitive exercises that target specific abilities/neural networks to potentially improve cognitive functioning through neuroplasticity. In a two-site study (NYSPI/Columbia and Duke) we will randomize 100 patients with MCI (broadly defined to include eMCI and lMCI as defined in the Alzheimer's Disease Neuroimaging Initiative, ADNI) to CCT (suite of exercises) or an active control condition (crossword puzzles). The initial 12 weeks will involve intensive CCT or active control conditions and this will be followed by regular booster sessions for a total of 18 months. We have extensive pilot data that strongly support this planned approach. We will address three key issues that have not been addressed systematically in a single well- controlled study in MCI: does CCT lead to improved cognitive functioning, do the effects of CCT transfer to functional ability and tasks of everyday life, and does CCT lead to long-term changes in brain networks, e.g., the default mode network (DMN) using resting fMRI? We aim to assess change in cognitive and functional status over 18 months in MCI patients comparing CCT versus active control condition, and hypothesize that MCI patients on active CCT will show better cognitive and functional outcomes by the end of the 18-month trial. We will also examine the sensitivity and validity of an unsupervised web-based neuropsychological battery (Brain Performance Test, BPT) for detecting CCT effects and its associations with change in cognitive and functional measures in MCI. We also hypothesize that patients with MCI randomized to CCT will show greater changes in the DMN with resting fMRI compared to the control condition and that these changes will be associated with improvement on specific cognitive tests. We will compare rates of transition to a clinical diagnosis of AD in the two groups, and explore potential moderators of improvement, including hippocampal and entorhinal cortex atrophy, olfactory identification deficits, and apolipoprotein E ?4 allele status. Improving cognition in older adults, especially those at high risk of transitioning to clinical AD, is an important public health goal. If the study goals are achieved it will have major implications for improving quality of life and reducing disability and healthcare costs in this growing demographic. This study is consistent with the NIA AD 2015 summit recommendations and NIA priorities, and the movement toward patient-driven personalized medicine.